I just noticed how awarded credits appear to be calculated. When three users return a result, each claims a certain amount of credit based on CPU time for the unit and their benchmark score. For example, say three users return a unit and claim 25, 40 and 100 credits, respectively, the middle value is awarded to all three users regardless, i.e. 40 credits. There doesn't appear to be any wieghting or averaging done, simply the middle value of the 3.

Also, the slower times for the seti 4.05 client do NOT adversely affect the credit requested. Simply, because the unit is taking longer to process, proportionally more credit is being claimed for it.

However, because linux for example still uses the older (faster) 4.02 seti client and the linux boinc client returns far lower benchmark scores, linux users are requesting far less credit for the same unit and this can significantly skew the awarded credit. For example, if a unit goes to 2 windows boxes and 1 linux box, because of the way credit is awarded, the Linux box will generally get more credit than requested. On the other hand, if a unit goes to 2 linux boxes and 1 windows box, then the windows box generally gets less credit than requested.

This senario doesn't appear uncommon (for linux users at least) because there appears to be far more windows boxes in use.

Edit: I don't have access to a Mac to compare requested credit against Windows and Linux boxes

Interesting stuff (if you care about credit!)

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

Awarding the middle result seems to be the case for some WU's however some have 3 results returned and are not getting any credit whatsoever like the one below.

http://setiweb.ssl.berkeley.edu/sah/workunit.php?wuid=1545811

---

> I just noticed how awarded credits appear to be calculated. When three users
> return a result, each claims a certain amount of credit based on CPU time for
> the unit and their benchmark score. For example, say three users return a unit
> and claim 25, 40 and 100 credits, respectively, the middle value is awarded to
> all three users regardless, i.e. 40 credits. There doesn't appear to be any
> wieghting or averaging done, simply the middle value of the 3.

There is a big problem if you start with weighting the results, someone that have figured out how to cheat and request a sh!tload of credits, like 5000000, would set the awarded credit waaaay higher than reasonable ... so I'm fine that htey use the median value. As now two persons must try to cheat on the result to have effect, and that is probably not likely to happen (as there's so many users and results)

---

________<br>We\'re not living in a black and white world

> > I just noticed how awarded credits appear to be calculated. When three
> users
> > return a result, each claims a certain amount of credit based on CPU time
> for
> > the unit and their benchmark score. For example, say three users return a
> unit
> > and claim 25, 40 and 100 credits, respectively, the middle value is
> awarded to
> > all three users regardless, i.e. 40 credits. There doesn't appear to be
> any
> > wieghting or averaging done, simply the middle value of the 3.
>
> There is a big problem if you start with weighting the results, someone that
> have figured out how to cheat and request a sh!tload of credits, like 5000000,
> would set the awarded credit waaaay higher than reasonable ... so I'm fine
> that htey use the median value. As now two persons must try to cheat on the
> result to have effect, and that is probably not likely to happen (as there's
> so many users and results)
>

Quite true :)

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

Ned,

> Interesting stuff (if you care about credit!)

Yes ...

This is however in the cosmetic class of problems ... :)

There is a general consensus that there are still issues with the benchmarks and the credit claiming (calculations). However, now is still not the time ... I know I sound like a broken record, but we have bigger fish to fry still ...

It is kinda like your car, a problem like "it won't start" is of slightly more importance than "my radio does not get my favorite station"...

If you need to get to work, which problem can you live with? :)

---